Improvement of accuracy of forest type classification and estimation of states of forest floor using multi-temporal satellite data based on phenological variation
Koide, Kaoru
This study is aimed at improving the accuracy of forest type classification by using remote sensing data as part of evaluating hydrological characteristics of the ground surface and subsurface for modelling regional groundwater flow. The study area is 5 km 5 km located at the Tono region, Gifu, central Japan. The forest in the study area is composed of three forest types; evergreen coniferous forest, deciduous broad-leaved forest and mixed forest consist of them. The author has attempted a forest type classification based on seasonal variation of NDVI calculated using multi-temporal SPOT data observed in summer, autumn and winter. The results of analysis realize that the distributions of residual NDVI between summer and autumn of coniferous forest and broad-leaved forest have good selectivity, the degree in separation of residual NDVI distributions between summer and winter is less than that between summer and autumn by influence of the forest floor evergreen plants, and the residual NDVI distribution between summer and winter is useful for classification between forest and grass field.